When we inhale, our lungs fill with oxygen, which is distributed into our red blood cells for transport throughout our body. Our bodies need a lot of oxygen to function, and a healthy person's oxygen saturation stays at 95%.
Respiratory diseases make it harder for the body to get oxygen from the lungs. This causes the oxygen saturation percentage to drop to 90% or lower, which indicates the need for medical attention.
In the clinic, doctors monitor blood oxygen saturation using pulse oximeters, those clips attached to fingertips or ears.
In a proof-of-principle study, researchers at the University of Washington and the University of California, San Diego showed that smartphones can detect blood oxygen saturation levels as low as 70 percent. This is the lowest value the U.S. Food and Drug Administration recommends that a pulse oximeter should be able to measure.
The technique involves participants placing their fingers on the camera and flash of a smartphone, which uses deep learning algorithms to decipher blood oxygen levels. When the team gave six subjects a controlled mixture of nitrogen and oxygen to artificially lower their blood oxygen levels, the smartphone correctly predicted whether the subjects had low blood oxygen levels 80 percent of the time.
Other smartphone apps that do so have been developed by asking people to hold their breath. But people can become very uncomfortable and have to breathe after about a minute, and that's before their blood oxygen levels drop enough to represent all clinically relevant data, with tests able to collect 15 from each subject minutes of data. The data shows that smartphones can work well within critical thresholds.
In this way, multiple measurements can be made with one's own equipment for free or at low cost, and ideally, the information can be seamlessly transferred to the doctor's office. This is great for telehealth appointments or triage nurses being able to quickly determine if a patient needs to go to the emergency room, or if they can continue to stay home and make an appointment with a primary care provider later.
To gather data to train and test the algorithm, the researchers had each participant wear a standard pulse oximeter on one finger, then place the other finger on the same hand over the smartphone's camera and flash . Each participant performed the same setup on both hands at the same time.
The camera is recording a video: as the heart beats, fresh blood flows through the part illuminated by the flash.
The camera records how much light from the flash is absorbed by the blood in the three color channels it measures: red, green and blue, and these intensity measurements can then be fed into a deep learning model.
Each participant inhaled a controlled mixture of oxygen and nitrogen to slowly lower oxygen levels. This process takes about 15 minutes. For all six participants, the team obtained more than 10,000 readings of blood oxygen levels between 61% and 100%.
The researchers used data from four participants to train a deep learning algorithm to extract blood oxygen levels. The rest of the data was used to validate the method, which was then tested to see how well it performed on new subjects.
Smartphone light is scattered by all these other components in the finger which means there is a lot of noise in the data we are looking at and deep learning is a very useful technique here because it can see these very complex and subtle features and helps you find patterns that you wouldn't otherwise be able to see.
One subject had thick calluses on their fingers, which made it harder for our algorithm to accurately determine their blood oxygen levels, and if we expanded this study to more subjects, we might see more People with calluses and more people of different skin tones. Then it's possible to have an algorithm that's complex enough to be able to better simulate all these differences.
The researchers say this is a good first step towards developing biomedical devices assisted by machine learning.
Facebook Conversations